I understand the logic behind it, having worked in a bar where, after pouring the 5-pints-2JD-coke-and-a-vodka-lemonade for the fifth time in the row, I actually felt like a robot. But I wonder if that robot would look and act like me, have we had a choice?
You might have heard the concept of ‘Uncanny Valley‘, but if you haven’t, here it is in a nutshell:
- If a machine behaves like a human we reject it
- Zombies are the scariest of creatures because they look human though they’re not
- The move it moves the less we expect it to look like a human
This may explain why certain robots look more like cute puppets than real people and why most restaurant robots don’t move a lot (if at all). But, these are also the very limitations of restaurant robots. Cruising up and down the tracks makes them exactly that – inhuman. But we go to restaurants to seek human contact. So what are our option but to make a robot that can walk around the tables and make conversations?
This is one of the issues NASA had to tackle head-on. After sending a few robots to planets, they realised that some of them, literally, get stuck in the mud. Not many options left but to teach it navigate, which, in modern terms, means using Artificial Intelligence. The aim was to make it smart enough to overcome obstacles and to look after itself. So, instead of programming the robot to go between certain, pre-determined destinations, they decided to make it find its own way. They did this by applying fuzzy logic and neural networks. In other words, make it think like a human would.
- Fuzzy logic is programmed through linguistics, not mathematics. The robot then basically thinks like a human would. If ask a person whether they can help you to move the table, they would start helping you instead of saying that its theoretically possible. Like in Nelly’s song… it’s getting hot in here so…
- Neural networks are, basically, a set of codes that allow a computer to learn over time. It’s using similar rules to the human brain. A computer like that learns like a baby, only a bit slower.
Using these two system allows the robot to learn about terrains, dangers and obstacles. The robots would learns the layout of the restaurants, that chairs are for sitting and tables are for eating, so that they know where to put the food. It means that we can move furniture around without having to re-programme the “staff” each time. It will also learn from people’s special requests, the most effective way to extract tips and even how to flirt.
We might find, that with a robot like that, we don’t need real people waiting tables. We might even have to make sure that they get paid a minimum wage. You never know, you might end up marrying one of them in the nearest future. Or at least make friends with one at your local coffee shop.